Health experts have raised concerns about the accuracy of some medical guidance provided by Google’s AI Overviews, following an investigation by The Guardian. The findings add to an ongoing debate about the reliability of AI-generated summaries in Google Search, particularly for health-related queries.
According to The Guardian, journalists tested a selection of medical searches and shared the resulting AI Overview summaries with medical charities, patient information groups, and health professionals. Several reviewers said the AI-generated responses contained misleading or incorrect information that could pose risks if followed.
How the Issues Were Identified
The investigation focused on how AI Overviews respond to common health queries. Medical organizations were asked to assess whether the guidance shown at the top of search results was accurate and safe.
In one example involving pancreatic cancer, Pancreatic Cancer UK said advice suggesting patients avoid high-fat foods was incorrect and potentially harmful. The charity warned that following such guidance could weaken patients and affect their ability to undergo treatment.
Mental health queries were also highlighted. Mind, a UK mental health charity, said some AI summaries relating to conditions such as psychosis and eating disorders included advice that was inaccurate, harmful, or could discourage people from seeking professional help.
In another case, cancer screening information was flagged as wrong. The Eve Appeal, a cancer charity, said an AI Overview incorrectly listed a pap test as a screening method for vaginal cancer.
Reviewers also noted that repeating the same search at different times could produce different AI-generated summaries, drawing from different sources. This variability made it difficult to consistently verify the information presented.
Google’s Response
Google disputed several aspects of The Guardian’s reporting. A company spokesperson said many of the examples shared were based on incomplete screenshots and did not reflect the full context of the AI Overviews.
Google stated that the vast majority of AI Overviews are factual and helpful, and that they link to well-known, reputable sources while encouraging users to seek expert medical advice. The company said it continuously makes quality improvements and that the accuracy of AI Overviews is comparable to other Search features, such as featured snippets.
Google also said that when AI Overviews misinterpret content or miss important context, it takes action in line with its policies.
Ongoing Concerns Around Accuracy
The investigation comes amid broader scrutiny of AI Overviews, which expanded in 2024. Early versions of the feature attracted criticism after producing clearly incorrect or nonsensical answers, prompting Google to limit the types of queries that trigger AI-generated summaries and refine how the system works.
Recent data suggests medical and YMYL (Your Money or Your Life) queries are more likely than average to trigger AI Overviews. Research has also raised questions about whether AI-generated answers are always fully supported by the sources they cite, even when links are provided.
Why This Matters
AI Overviews appear above traditional search results, giving them added visibility and authority. When inaccuracies appear in health-related summaries, the potential impact is higher than in other categories.
The findings also highlight a practical challenge: because AI Overviews can change between searches, users and publishers may struggle to reproduce or verify the exact information shown at a given time.
What Comes Next
Google has previously adjusted AI Overviews in response to public criticism and says it will continue to improve the feature. Its response to The Guardian suggests the company views AI Overviews as part of its broader Search experience, rather than as a separate product held to a different standard.
How Google balances speed, visibility, and accuracy, particularly for sensitive health queries, will likely remain a key focus as AI Overviews continue to evolve.
You Might Also Like: New Study Reveals 37% of Consumers Are Turning to AI Before Google